Google Expands Footprint in Defense With New AI Pentagon Deal

In Focus

• Google will supply AI models for classified work to the Pentagon
• The DoW will use Google’s AI models for any lawful purpose
• Google AI systems will not be used for domestic surveillance

Google has inked an agreement with the U.S. Department of War (DoW) to use its AI models for classified work. A report by The Information shows that under the Google Pentagon AI deal, the DoW will use the AI models for “any lawful government purpose”.

Strong Focus on AI Control and Oversight

Google’s AI defense contract with the Pentagon reportedly requires the tech giant to adjust AI safety settings and filters upon the government’s request. It however states that the government would not use AI systems for domestic surveillance.

“The parties agree that the AI System is not intended for, and should not be used for, domestic mass surveillance or autonomous weapons (including ​target selection) ​without appropriate human oversight and control,” the agreement reads in part as cited by Reuters.

Google asserted its belief that AI systems should not be used for domestic mass surveillance or autonomous weaponry without the necessary human control and oversight.

“We believe that providing API access to our ​commercial models, including on Google infrastructure, with ​industry-standard practices and terms, represents a responsible approach to supporting national security,” A Spokesperson for Google noted.

Google supports federal agencies in both classified and unclassified work. Last year, the Department of War deployed a Gemini-powered AI platform on GenAI.mil to provide military personnel with access to advanced AI capabilities. The platform supports unclassified defense workflows.

Pentagon Emphasizes Lawful Use of AI

The Pentagon emphasized that Google should allow ‘any lawful use’ of its AI systems by the government. The latest agreement limits Google’s right to control or veto lawful operational decision-making by the government.

The defense agency said it had no intention of using artificial intelligence to surveil Americans or build weapons that operate without human control. Unrestricted AI use by the military was at the heart of the fallout between Anthropic and the Pentagon earlier this year. This fallout triggered a conversation about the use of AI models for national security.

Early last month, strong public backlash pushed OpenAI into amending deal with the PentagonI to explicitly prohibit the use of AI systems for domestic surveillance on U.S. citizens or nationals.

Google Joins OpenAI in Supplying AI Systems

Following the AI classified contract, Google joins tech companies such as OpenAI and xAI, which have also signed agreements to supply AI models for classified use. The Pentagon uses classified networks to handle a range of sensitive tasks, which include planning missions and targeting weapons.

In 2025, the Pentagon signed multiple deals, each valued up to $200 million, with leading AI firms including OpenAI, Anthropic, and Google. A part of the U.S. defense AI strategy, the agency has reportedly been encouraging AI developers to make their tools available on classified networks. The Pentagon wants AI firms to do so without the standard restrictions that are applicable to other users.

Last month, OpenAI started selling AI models to U.S. federal agencies for both classified and unclassified work. Previously, the AI developer focused on supplying AI for unclassified government use. But this changed after a recent deal with the Pentagon paved the way for OpenAI to supply models for classified operations.